Long-Term Seizure Outcomes in Autoimmune Encephalitis

Abstract

Introduction

Seizures are common in autoimmune encephalitis (AE), but identifying patients at risk of chronic epilepsy in the post-acute phase remains challenging. This study aims to identify risk factors of treatment-resistant postencephalitic epilepsy.

Methods

This retrospective cohort study included patients with AE who experienced new-onset seizures within one year of symptom onset from two tertiary care centers in New York. EEG findings were analyzed separately based on whether the EEG recording was obtained in the acute (<3 months from symptom onset) or subacute phase. A multivariate logistic regression model was used to identify independent predictors of postencephalitic epilepsy.

Results

Eighty-nine patients were included (median age: 33 years). Neural antibodies were present in 73% of patients (NMDAR: 35, LGI1: 19, GAD65: 9, Hu: 1, AGNA-1: 1). Over a median follow-up of 4.9 years, 29.2% developed treatment-resistant postencephalitic epilepsy. Independent predictors of postencephalitic epilepsy included focal slowing on acute EEG (OR 0.15, CI 0.02-0.90), interictal epileptiform discharges (IEDs) or periodic discharges (PDs) on subacute EEG (OR 20.01, CI 1.94-206.44), and cell surface antibodies (OR 0.21, CI 0.05-0.89). Immunotherapy within three months of onset was associated with decreased epilepsy development in patients with neural antibodies (OR 4.16, CI 1.11-16.30).

Conclusions

Nearly one-third of patients with AE and acute seizures developed treatment-resistant postencephalitic epilepsy, with significant predictors including absence of focal slowing on acute EEG, presence of IEDs and PDs on subacute EEG, absence of cell surface antibodies, and absence of early immunotherapy treatment of patients with positive neural antibodies.

Introduction

Seizures are a common manifestation of autoimmune encephalitis (AE), with approximately 70% of patients presenting with seizures during their illness. ¹ They can be conceptually classified into two categories as defined by the International League Against Epilepsy (ILAE) Autoimmunity and Inflammation Taskforce: acute symptomatic seizures, which occur in the active phase of AE, and autoimmune-associated epilepsy (AAE), which leads to chronic seizures and is associated with structural injury. While patients with acute symptomatic seizures typically respond well to immunotherapy, those that develop autoimmune-associated epilepsy may be more resistant to both antiseizure medications (ASMs) and immunotherapy. ²

It is unclear which factors predict long-term seizure outcomes after recovery from AE. Several studies have shown that the risk of chronic epilepsy is higher in patients with AE associated with intracellular antibodies, such as glutamic acid decarboxylase 65 (GAD65) antibodies, compared to neuronal surface antibodies, such as NMDA receptor (NMDAR) and leucine-rich glioma-inactivated 1 (LGI1) antibodies.^3-6 Delay in immunotherapy initiation during initial presentation has also been consistently found to be associated with development of chronic seizures.^4,7-10 However, other factors remain uncertain or have been inconsistently reported as significant, including age of onset; presence of specific EEG patterns, including interictal epileptiform discharges (IED); structural features on MRI; and type of immunotherapy used.^7,11-19 Though some ASMs have been found to act additively to immunotherapy, such as carbamazepine and valproate, it is also unknown if the type of ASM initiated affects long-term seizure outcomes.^20,21 In addition, there are relatively few reported long-term outcomes of antibody negative cases, and it is unclear if there are differences in prognostic factors between antibody negative and positive cases.^4,7,11,16,22

This study aims to identify predictive factors associated with the development of postencephalitic epilepsy after recovery from AE, examining clinical and paraclinical factors in a diverse patient population across two New York City tertiary care centers, using comprehensive EEG data from both the acute and subacute phases of the disease. We hypothesize that EEG, a sensitive tool for the identification of patients with AE, ¹ has prognostic value that is modulated by the time from symptomatic onset to study acquisition. This study includes patients both with and without neural antibodies to reflect the full breadth of AE syndromes encountered in the clinic and to assess for differences in seizure outcomes and response to treatment between these subgroups.

Methods

Study Design and Patient Population

This is a retrospective cohort study of patients from Columbia University Irving Medical Center (CUIMC) and New York University (NYU) Langone Medical Center treated for AE who experienced new-onset seizures within one year from symptom onset. The CUIMC cohort was identified through an ICD-9 and ICD-10 code search of the electronic medical records for encephalitis diagnoses in patients of all ages seen at CUIMC between January 1, 2000 and December 31, 2020, followed by manual chart review for diagnostic confirmation. The NYU cohort was identified through an ICD-10 code search for encephalitis diagnoses in adult patients seen at NYU between January 1, 2012 and December 31, 2017, then by medical practice and notification of hospitalized patients to study investigator CS from January 1, 2018 to January 31, 2023, followed by manual chart review. Of note, acute seizure outcomes of 34 of the included patients from CUIMC are presented in a separate presentation on acute symptomatic seizure outcomes, ²³ which focus on short-term outcomes during initial hospitalization compared to the more longitudinal outcomes presented here.

The study included patients with positive neural antibodies (NMDAR, LGI1, CASPR2, GABA-A, GABA-B, GFAP, amphiphysin, Hu, Yo, Ri, Ma, AGNA-1, and high titer GAD65 with serum titers >20 nmol/L or >200 IU/mL) with clinical presentation consistent with AE, or patients meeting Graus criteria for definitive limbic encephalitis or probable antibody negative AE. ²⁴ We excluded patients with a history of seizures prior to AE symptom onset, insufficient documentation of diagnostic workup and treatment course during the acute phase of the disease, and those with less than one year of follow-up after initial symptom onset.

Predictive Variables

We reviewed medical records for patient demographics, socioeconomic factors, clinical characteristics, acute and subacute EEG findings, brain MRI, CSF profile, malignancy screening, and ASM and immunotherapy treatment course. EEGs were assessed for background slowing, focal slowing, rhythmic delta activity (RDA), interictal discharges (IED), periodic discharges (PD), and electrographic seizures according to the American Clinical Neurophysiology Society 2021 guidelines. ²⁵ The acute phase was defined as within three months from initial AE symptom onset, and the subacute phase was defined as three to 12 months after initial AE symptom onset. Study data were collected and managed using REDCap electronic data capture tools hosted at CUIMC and NYU.

Outcomes

The primary outcome measured was treatment-resistant postencephalitic epilepsy, defined as the presence of any clinical or electrographic seizures within the past six months at the last follow-up date. While there is no formal definition on post-encephalitic epilepsy, we chose this time point based on current expert opinion and knowledge on the acute phase of acute inflammation in AE. ² Secondary outcomes reported include time to seizure freedom, ASM discontinuation, time to ASM freedom, rehospitalization for AE symptoms after initial hospitalization, changes in seizure semiology, and relapse. Relapse was defined as the new-onset or worsening of symptoms after at least three months of clinical improvement or stabilization.

Statistical Analysis

Descriptive analysis was performed with medians and interquartile ranges for continuous variables and proportions for categorical variables. To identify possible bias from loss to follow-up, we assessed for any differences in sociodemographic or clinical characteristics between the patients included in our study and those with insufficient follow-up. Univariate analysis was performed using Wilcoxon rank-sum test for continuous variables and Fisher’s exact test for categorical variables. Multivariate logistic regression with backward selection based on AIC (Akaike Information Criterion) including age, sex, and variables with P < 0.05 on univariate analysis was applied to identify predictive factors of treatment-resistant post-encephalitic epilepsy. To handle missing data in covariates, multiple imputation with ten imputed datasets was performed. Variables that were selected at least once on imputed datasets were included in the final model. Rubin’s rule was applied to calculate standard errors. Statistical significance was defined as P < 0.05.

All statistical analysis was done using R statistics (Version 4.3.1, 2023, The R Foundation for Statistical Computing, Vienna, Austria).

Standard Protocol Approvals, Registrations, and Patient Consents

This study was approved by the CUIMC and NYU Research Ethics Boards, with a waiver of informed consent obtained for a retrospective review.

Data Availability

Deidentified data not published in this article will be made available by request of any qualified investigator for purposes of replicating procedures and results.

Results

Sociodemographic and Clinical Characteristics

We identified 111 patients with AE and new-onset seizures within one year of AE symptom onset with sufficient documentation of workup and treatment received during the acute and subacute phase of disease. Of those, 89 (80.2%) had at least one year of documented neurologic follow-up and were included in analysis (Figure 1). The 22 patients with insufficient follow-up either died within the first year of disease (n = 6) or were lost to follow-up (n = 16). Except for older age, these patients were not significantly different in sociodemographic or clinical characteristics compared to those with sufficient follow-up (Supplemental Table 1).

Figure 1.

Study flowchart. ^a If workup was done during the acute phase of disease, we required available documentation of EEG, MRI, and CSF profile results; type of immunotherapy and ASM treatments; and documentation of whether workup and treatment was performed in the acute or subacute phase. AE: autoimmune encephalitis; ASM: antiseizure medication CUIMC: Columbia University Irving Medical Center; NYU: New York University.

For the 89 study participants included, median age at disease onset was 33 years (IQR 37). Study participants were diverse in ethnicity, country of origin, and insurance status (Table 1). Neural antibodies were found in 65 (73.0%) patients, including NMDAR (n = 35), LGI1 (n = 19), GAD65 (n = 9), Hu (n = 1), GABA-B (n = 1), and AGNA-1 (n = 1). One patient was positive for both NMDA and GABA-B antibodies. Nearly all patients (n = 87, 97.8%) experienced clinical seizures (two patients experienced electrographic only seizures), and most of these patients experienced their first seizure within three months of initial symptom onset (n = 83, 95.4%) Focal to bilateral tonic clonic (61, 70.1%) were the most common seizure type, followed by focal with impaired awareness seizures (42, 48.3%) (Table 1).

Table 1.

Study Patient Demographic and Clinical Characteristics.

Patient Characteristics	n = 89
Demographics
Age, median (IQR)	33 (37) years
<18 years	18 (20.2%)
Sex
Male	38 (42.7%)
Female	51 (57.3%)
Race
White	39 (43.8%)
Other	20 (22.5%)
Black or African American	17 (19.1%)
Asian	12 (13.5%)
Unknown	1 (1.1%)
Ethnicity
Hispanic/Latino	15 (16.9%)
Non-Hispanic/Latino	60 (67.4%)
Unknown	14 (15.7%)
English proficiency
Proficient	81 (91.0%)
Limited	8 (9.0%)
Country of origin
United States	19 (21.3%)
Other a	21 (23.6%)
Unknown	49 (55.1%)
Socioeconomics
Employment
Working	40 (44.9%)
In school	22 (24.7%)
Retired	9 (10.1%)
Unemployed	10 (11.2%)
Unknown	5 (5.6%)
Not applicable (<5 years old)	5 (5.6%)
Insurance
Private/Commercial insurance	66 (74.2%)
Medicare	11 (12.4%)
Medicaid	19 (21.3%)
Uninsured	3 (3.4%)
Antibody type
Autoantibody negative	24 (27.0%)
Autoantibody positive b	65 (73.0%)
NMDAR	35 (53.8%)
LGI-1	19 (29.2%)
GAD65	9 (13.8%)
Hu	1 (1.5%)
GABA-B	1 (1.5%)
AGNA-1	1 (1.5%)
Cell surface Ab	54 (83.1%)
Intracellular Ab	11 (16.9%)
Clinical characteristics	n = 89
Initial hospitalization c
Hospitalization during acute phase	78 (87.6%)
Length of hospitalization, median (IQR)	29.5 (38.8) days
Time to admission, median (IQR)	6 (17) days
	n = 78
Transferred from another hospital	42 (53.8%)
ICU Admission	44 (56.4%)
Intubation	31 (39.7%)
Discharge location
Home	34 (43.6%)
Rehab	35 (44.9%)
Hospital	5 (6.4%)
Nursing home	3 (3.8%)
Neurological symptoms (acute phase)	n = 89
Altered mental status	63 (70.8%)
Psychiatric symptoms d	33 (37.1%)
Memory deficit	30 (33.7%)
Hallucinations	15 (16.9%)
Headache	31 (34.8%)
Generalized weakness	2 (2.2%)
Focal weakness	6 (6.7%)
Speech changes	12 (13.5%)
Clinical seizures	83 (93.3%)
Clinical seizure characteristics	n = 87
Onset 0-3 months	83 (95.4%)
Onset 3-12 months	4 (4.6%)
Time from initial symptom to seizure onset, median (IQR)	2 (10) days
Seizure type
FBDS	11 (12.6%)
Focal aware	21 (24.1%)
Focal with impaired awareness	42 (48.3%)
Focal to bilateral tonic clonic	61 (70.1%)
Unknown	1 (1.1%)
>1 seizure type	42 (48.3%)
Maximum clinical seizure frequency
Hourly	28 (32.2%)
Daily	16 (18.4%)
Weekly	9 (10.3%)
Monthly	11 (12.6%)
Less than monthly	19 (21.8%)
Once	4 (4.6%)
Clinical status epilepticus	28 (32.2%)

^aCaribbean (9), South America (4), Asia (3), Europe (3), Africa (1), unknown (1).

^b1 patient with both NMDAR and GABA-B antibodies.

^cIf a patient had multiple hospitalizations during the acute phase, only data from first admission was considered. If a patient was discharged and readmitted within 7 days, this was considered as the same admission.

^dAnxiety, depression, psychosis.

Ab: antibody; AE: autoimmune encephalitis; AGNA-1: antiglial/neuronal nuclear autoantibody-type 1; FBDS: faciobrachial dystonic seizure; GABA-B: gamma-aminobutyric acid B; GAD65: glutamic acid decarboxylase 65-kilodalton isoform; ICU: intensive care unit; IQR: interquartile range; LGI-1: leucine-rich glioma-inactivated 1; NMDAR: N-methyl-d-aspartate receptor.

EEG and Other Diagnostic Testing

Seventy of the 82 patients (85.4%) who had an EEG done in the acute phase had abnormal findings, while 48 of 58 patients (82.8%) who had an EEG in the subacute phase had abnormalities (Supplemental Table 2). Eighty-four patients (94.4%) had a brain MRI done in the acute phase and 46 (51.7%) had a brain MRI done in the subacute phase, with median time to first MRI of 11 days (IQR 21.5). Most patients who underwent lumbar puncture in the acute phase were found to have CSF pleocytosis (n = 50, 71.4%), while only three (15.0%) of the patients who underwent lumbar puncture in the subacute phase had pleocytosis. Brain biopsy was performed in few patients (n = 9), with two-thirds of biopsied patients with evidence of inflammation on pathology (Table 2).

Table 2.

Diagnostic Results and Immunotherapy Treatment.

Diagnostics	n = 89
EEG a
Acute EEG performed	82 (92.1%)
Any acute EEG abnormality	70 (85.4%)
Background slowing	52 (63.4%)
Focal slowing	51 (62.2%)
Rhythmic delta activity	32 (39.0%)
Interictal epileptiform discharges	36 (43.9%)
Periodic discharges	17 (20.7%)
Electrographic seizures	44 (53.7%)
Subacute EEG performed	58 (65.2%)
Any subacute EEG abnormality	48 (82.8%)
Background slowing	28 (48.3%)
Focal slowing	32 (55.1%)
Rhythmic delta activity	10 (17.2%)
Interictal epileptiform discharges	21 (36.2%)
Periodic discharges	3 (5.2%)
Electrographic seizures	18 (31.0%)
Initial negative → positive EEG findings	12 (13.5%)
MRI Brain
Acute MRI performed	84 (94.4%)
Acute MRI abnormality	53 (63.1%)
Subacute MRI performed	46 (51.7%)
Subacute MRI abnormality	31 (67.4%)
Localization of MRI abnormality (acute or subacute phase) b
Temporal	48 (53.9%)
Frontal	20 (22.5%)
Parietal	11 (12.4%)
Occipital	6 (6.7%)
Basal ganglia	7 (7.9%)
>1 affected area	26 (29.2%)
Initial negative → positive MRI findings	24 (26.7%)
CSF results
Acute LP performed	70 (78.7%)
Acute pleocytosis	50 (71.4%)
Subacute LP performed	20 (22.5%)
Subacute pleocytosis	3 (15.0%)
Maximum WBC count, median (IQR)	28 (32) WBC/mm3
CSF Oligoclonal band testing performed	51 (57.3%)
Positive oligoclonal bands	24 (47.1%)
Initial negative → positive CSF pleocytosis	5 (5.6%)
Other diagnostic testing
Brain biopsy performed	9 (10.1%)
Evidence of inflammation	6 (66.7%)
Associated malignancy c	15 (16.9%)
Time to diagnostic antibody testing result, median (IQR)	50 (111.3) days
Immunotherapy	n = 89
Any immunotherapy	87 (97.8%)
Time to any immunotherapy, median (IQR)	24.5 (75.3) days
Any immunotherapy in acute phase	67 (75.3%)
Steroids	83 (93.3%)
Time to steroid initiation, median (IQR)	27 (61.5) days
Steroids in acute phase	66 (74.2%)
>1 IVMP course	29 (32.6%)
IVIG	72 (80.9%)
Time to IVIG, median (IQR))	33 (74.8) days
IVIG in acute phase	54 (60.7%)
>1 IVIG course	44 (49.4%)
PLEX	26 (29.2%)
Time to PLEX, median (IQR)	28 (61) days
PLEX in acute phase	20 (22.5%)
Other immunotherapy	56 (62.9%)
Rituximab	44 (49.4%)
Time to rituximab, median (IQR)	93 (339.3) days
Rituximab in acute phase	21 (23.6%)
>1 rituximab course	35 (39.3%)
Cyclophosphamide	8 (9.0%)
Time to cyclophosphamide, median (IQR)	59 (284.8) days
Cyclophosphamide in acute phase	5 (5.6%)
Mycophenolate mofetil	16 (18.0%)
Time to MMF, median (IQR)	367.5 (673.8) days
MMF in acute phase	3 (3.4%)
Azathioprine d	4 (4.5%)
Tocilizumab d	2 (2.2%)
Anakinra d	1 (1.1%)

Percentages calculated from number of patients who received the diagnostic test during the acute or subacute phase.

^aMore detailed EEG findings described in Supplemental Table 2.

^bIn either acute or subacute phase.

^cAssociated malignancies include ovarian teratoma, thymoma, small cell lung carcinoma, ovarian serous cystadenofibroma, mantle cell lymphoma, multiple myeloma, melanoma, and gastrointestinal stromal tumor.

^dInitiated in post-acute phase of disease.

AE: autoimmune encephalitis; IQR: interquartile range; IVIG: intravenous immune globulin; IVMP: intravenous methylprednisolone; LP: lumbar puncture; MMF: mycophenolate mofetil; PLEX: plasmapheresis; WBC: white blood cell.

Treatment

Most patients were initiated on an ASM during the acute phase of disease (n = 81, 91.0%). The most common ASM used during the acute phase was levetiracetam (n = 74) followed by phenytoin (n = 37) and lacosamide (n = 28), with a median of two ASMs (IQR 3) started during this time. Median time from seizure onset to ASM initiation was one day (IQR 10). More than one-third of patients discontinued ASMs due to side effects at some point in their treatment course, with levetiracetam as the most common offending agent (n = 14, 15.7%), largely due to psychiatric side effects (Supplemental Table 3). Nearly all patients received some form of immunotherapy during their disease (97.8%), most with initiation in the acute phase (75.3%), with median time to immunotherapy initiation from symptom onset of 24.5 days (IQR 75.3) (Table 2).

Outcomes

Over a median and mean follow-up of 4.9 years (IQR 4.8) and 5.7 years (range: 1.1-18.5), respectively, 26 (29.2%) patients developed postencephalitic epilepsy. For those that did not develop epilepsy, median and mean time to seizure freedom was 3.1 months (IQR 16.7) and 18.5 months (range: 0-129.6). Only 37 (41.6%) patients discontinued all ASMs at their last follow-up date, with median and mean time to ASM freedom of 20.0 months (IQR 20.4) and 27.3 months (range: 1.0-142.8). 33 (37.1%) patients achieved both seizure and ASM freedom. Patients with persistent seizures (n = 26, 29.2%) experienced seizures with varying frequencies, with 14 patients (53.8%) experiencing seizures monthly or less, 4 (15.4%) experiencing seizures weekly, and 7 (26.9%) experiencing seizures daily or more. At last follow-up, 51 (57.3%) study participants had experienced rehospitalization for AE symptoms after their initial hospitalization, two (2.2%) were deceased, and 29 (32.6%) had experienced relapse, with a median number of one relapse (IQR 1) and median time to first relapse of 14.7 months (IQR 25.8). 86.2% (n = 25) of patients who experienced AE relapses had symptoms associated with recurrence or increased frequency of seizures. Twenty-three (25.8%) patients developed seizures with new semiology after their initial seizure presentation, most commonly new focal aware seizures (n = 13), followed by focal with impaired awareness (n = 5), focal to bilateral tonic clonic (n = 3), and faciobrachial dystonic seizures (n = 1).

When comparing outcomes between antibody positive and negative patients, we found no significant differences in the proportion of patients who experienced seizure freedom, rehospitalization, or relapse. However, while 52.3% of antibody positive patients discontinued all ASMs, this only occurred in 12.5% of antibody negative patients (P < 0.001) (Figure 2A).

Figure 2.

Antibody positive vs negative patient response to acute immunotherapy and long-term seizure outcomes. *P < 0.05, **P < 0.01. ASM: antiseizure medication; IVIG: intravenous immune globulin; PLEX: plasmapheresis.

Predictive Factors

On univariate analysis, intracellular antibodies, temporal abnormalities on MRI, and IEDs or PDs on subacute EEGs (obtained at 3-12 months from initial symptom onset) were found to be significantly associated with the development of treatment-resistant postencephalitic epilepsy. In contrast, presence of cell surface antibodies, altered mental status, focal to bilateral tonic clonic seizures, focal slowing on acute EEGs (obtained within 3 months of initial symptom onset), steroids in the acute phase, and IVIG in the acute phase were found to be protective against the development of epilepsy. From the final multivariate logistic regression model, we found that IEDs/PDs on subacute EEG were independent significant predictors of development of treatment-resistant postencephalitic epilepsy (OR 20.01, CI 1.94-206.44), and focal slowing on acute EEG (OR 0.15, CI 0.02-0.90) as well as presence of cell surface antibodies (OR 0.21, CI 0.05-0.89) were inversely associated with epilepsy (Table 3). Of the 21 patients with IEDs/PDs on subacute EEG, 8 patients also had IEDs/PDs on their acute EEG while 8 patients did not (the remaining 5 patients did not have an EEG available in the acute phase). 6/8 patients with new IEDs/PDs on subacute EEG developed epilepsy, while 3/8 patients with IEDs/PDs already present on acute EEG developed epilepsy; changes in findings of IEDs/PDs on EEG from the acute to subacute phase did not significantly predict the development of epilepsy (P = 0.31).

Table 3.

Logistic Regression for Factors Associated With Development of Treatment-Resistant Postencephalitic Epilepsy.

	Unadjusted a			Adjusted
	OR	95% CI	P	OR	95% CI	P
Age	1.01	0.98-1.03	0.64	0.99	0.95-1.03	0.64
Male sex	0.98	0.39-2.46	0.96		Not selected
Acute EEG focal slowing	0.36	0.13-0.98	0.05	0.15	0.02-0.90	0.04
Subacute EEG IED/PD	7.74	2.34-25.58	0.002	20.01	1.94-206.44	0.02
Tonic clonic seizure semiology	0.31	0.12-0.82	0.02	0.24	0.05-1.08	0.07
Altered mental status	0.20	0.07-0.55	0.002	0.26	0.06-1.19	0.09
Cell surface antibodies	0.33	0.11-0.99	0.05	0.21	0.05-0.89	0.04
Intracellular antibodies	4.44	0.93-21.22	0.06	2.32	0.18-29.87	0.52
Temporal MRI abnormalities	3.12	1.15-8.49	0.03		Not selected
Acute steroids	0.32	0.12-0.87	0.03		Not selected
Acute IVIG	0.27	0.10-0.70	0.009	2.43	0.27-21.66	0.43

^aValues obtained from multiple imputed dataset. CI: confidence interval; IED: interictal epileptiform discharge; IVIG: intravenous immune globulin; OR: odds ratio; PD: periodic discharge.

Further analysis of EEG descriptors showed that although presence of acute focal slowing was associated with seizure freedom, within the subgroup of patients who did experience acute focal slowing, increased prevalence on EEG, defined by the proportion of the record that included focal slowing, was associated with persistent seizures (eg, patients with frequent focal slowing were more likely to develop epilepsy than those with rare focal slowing) (P = 0.037). Patients with unilateral RDA also had a greater chance of developing persistent seizures than those with either generalized or independent bilateral RDA (P = 0.027) (Supplemental Table 4). Presence of electrographic seizures during either the acute or subacute phase was not significantly associated with the development of epilepsy; of the 10 patients who experienced subacute electrographic seizures and ultimately achieved seizure freedom, two had LGI1 antibodies, six had NMDAR antibodies, and two were autoantibody negative.

No statistically significant differences in the development of treatment-resistant postencephalitic epilepsy were found based on the measured sociodemographic factors, length of hospitalization, ICU admission, intubation, maximum seizure frequency, clinical or electrographic status epilepticus, CSF findings, or malignancy association. No particular ASM initiated in the acute phase nor ASM class including sodium channel blocking agents (NCBs) or GABAergic agents were found to be associated with differences in postencephalitic epilepsy outcome, nor time to ASM initiation or total number of ASMs used (Supplemental Table 5). Subgroup analysis of LGI1 patients did not show differences in outcome based on exposure to NCBs, and no differences in response to specific ASMs were seen between antibody positive and negative patients. In general, acute immunotherapy of any type was associated with decreased epilepsy development, though no particular therapy was found to be significant on multivariate analysis. However, when separately analyzing patients with (n = 65) and without (n = 24) neural antibodies, acute steroids (P = 0.027) and IVIG (P = 0.019) had an association with decreased development of epilepsy only in those with neural antibodies, whereas patients with antibody negative encephalitis fared similarly with or without first-line immunotherapy. (Figure 2B).

Discussion

In this retrospective cohort study of 89 patients with AE and new-onset seizures, we report the long-term seizure outcomes of a diverse patient population across two New York City tertiary care centers. In our study, nearly one-third of patients developed treatment-resistant postencephalitic epilepsy, which is slightly lower than previous studies including patients both with and without neural autoantibodies that reported rates of persistent seizures from 39% to 58%.^4,7,11,16,20 Approximately 70% of patients with AE and seizures were seizure free at long-term follow up and at least 45% remained seizure free off ASM. IEDs or PDs on EEGs obtained between 3 to 12 months after symptom onset were found to be independent predictors for the development of epilepsy, while focal slowing on EEGs obtained during the acute phase of AE and cell surface antibodies were found to be predictors against the development of epilepsy.

Interestingly, IEDs and PDs on subacute but not acute EEG were found to be a predictor of epilepsy. Earlier studies had shown conflicting evidence on whether presence of IEDs on EEG is predictive of long-term seizure outcomes,^7,10-14 and our findings suggest that timing of EEG analysis in the patient’s disease course may be a contributing factor. Indeed, in one study, ⁴ persisting IEDs on follow-up were found to be a risk factor for the development of epilepsy, supporting the idea that follow-up EEGs after the initial acute phase of disease may be more important for prognostic value. These findings also align with other studies demonstrating the predictive value of IEDs in the development of epilepsy following acute brain injuries such as stroke and complex febrile seizures.^26,27 Of note, presence of electrographic seizures during the subacute phase was not significantly associated with epilepsy, demonstrating that persistence of seizures at this stage does not preclude eventual seizure freedom and may be due to an ongoing active encephalitic process.

In contrast, the development of epilepsy was inversely associated with presence of cell surface antibodies, as had been reported in previous studies,^3-6 as well as focal slowing on acute EEG. Several predictive factors of seizure outcomes in prior studies such as sex, ¹⁷ age,^14,17,19 seizure frequency,^11,22 status epilepticus,^8,14,22 and memory deficit ¹¹ were not found to be significant in our study. These discrepancies in findings may be due to differences in the included patient population. Altered mental status and focal to bilateral tonic-clonic seizures were inversely associated with post-encephalitic epilepsy in our cohort but this may also be due patient population differences. While most prior studies reported on samples selected based on a single antibody subtype, we included patients with all neural antibody types and antibody negative patients, as well as both pediatric and adult patients. The demographic composition of our study population may also differ, especially compared to studies with different patient populations.

In our study, no differences in epilepsy outcome were found based on ASMs started in the acute phase, including response to NCBs in patients with LGI1 antibodies as had been seen in some studies..^28,29 In addition, there were no apparent differences in response to any particular ASM between antibody positive and antibody negative patients, though exact comparisons were difficult to make because patients were started on ASMs in different orders, often concurrently, and for different lengths of time. A relatively large proportion of patients were found to discontinue an ASM at some point in their disease course due to side effects; in particular, levetiracetam was discontinued in a considerable number of patients due to psychiatric effects, supporting previous suggestions that it may not ultimately be the best ASM to start in patients with AE due to its side effect profile. ²⁸

We were also able to make comparisons between cases where patients did and did not receive immunotherapy within the first three months of symptom onset due to the extensive duration of our database, which spans years when treatment initiation was less aggressive, as well as cases with delays to initial health care consultation. In contrast to ASMs, immunotherapy initiated in the acute phase was inversely associated with treatment-resistant postencephalitic epilepsy, and though it did not achieve statistical significance on multivariate analysis, this finding is consistent with previous literature on the importance of early initiation of immunotherapy in this patient population.^4,7-10 Interestingly, this effect was seen only in neural antibody positive patients. While the lack of significant effect in antibody negative patients may be due to lack of power given the small sample size of the subgroup, it may also indicate that first-line immunotherapy is not sufficient to treat this patient population, possibly due to a different pathogenic mechanism of disease. ³⁰ Future studies investigating immunotherapy in antibody negative patients may be needed to assess for potential differences in response to treatment.

There are also relatively few studies that directly compare outcomes between antibody positive and antibody negative patients, with conflicting evidence on whether they significantly differ in seizure outcome.^4,11,22 In our study, we found that while these patients did not differ in the primary outcome of treatment-resistant postencephalitic epilepsy development, antibody negative patients were significantly less likely to discontinue ASMs. This may suggest overall worse outcomes in antibody negative patients, though it is difficult to fully assess whether these patients were justified in remaining on ASMs. It may also reflect a more conservative management of these patients, possibly due to the limited long-term seizure outcome data available for this patient subset.

Limitations to our study include its retrospective design, antibody heterogeneity, and smaller sample sizes within specific antibody groups, making it difficult to conduct subgroup analyses. We felt that including the full variety of AE types was valuable to reflect real-life practice, especially since patients may face delays to confirmatory antibody testing and many patients are never found to have any neural antibodies. Another limitation concerns the timeframe of seizure onset as we chose six months after symptom onset for primary outcomes of seizure freedom. This is a potentially arbitrary timeframe as current expert opinion does not provide a timeline of postencephalitic epilepsy development, which typically occurs without a latent period. The wide spectrum of clinical presentation can vary according to the particular associated antibody and timing of immune-targeted therapy and requires ancillary evidence of persistent encephalitis. Therefore, we recognize that this may not accurately assess the burden of post-encephalitic epilepsy. Further studies are needed to evaluate the most appropriate time period of measurement. In addition, while our study included a multi-ethnic patient cohort, it was conducted in tertiary care centers, which may capture a sicker patient population and thus limit the generalizability of our results to patients seen in other settings.

In conclusion, our study provides insight on the clinical course and long-term seizure outcomes of a diverse cohort of patients with AE. Treatment-resistant postencephalitic epilepsy is a common outcome in many of these patients and can be predicted by clinical and paraclinical findings from both the acute and subacute phases of disease, including presence of IEDs and PDs on subacute EEG, absence of focal slowing on acute EEG, and absence of cell surface antibodies. The identification of these independent predictors of epilepsy may help guide clinical decision making, including risk-stratification of patients with AE to determine the appropriate timing of discontinuation of ASM use and need for long-term monitoring, ultimately reducing unnecessary prolonged ASM use in lower-risk patients. Further prospective studies are warranted to validate these findings and to further study the potential differential response to immunotherapy based on antibody status.

ORCID iDs

Carla Y. Kim https://orcid.org/0000-0001-5714-7212

Runze Cui https://orcid.org/0009-0008-8684-6686

Kiran T. Thakur https://orcid.org/0000-0003-0050-0323